<< | [up] | >> |
dim_STAT User's Guide. by Dimitri |
Start On-Line Collecting |
Before starting any STAT collect, first check if the STAT-service is running on every server you want to monitor. This is the most common error!!Another point, if you want to monitor a Linux server, be sure you've installed the Linux STAT Add-Ons, before starting any collect (see the special Linux section in this document).
Now, from the dim_STAT Main Page you may just follow the Start New Collect link. (Note: since version 8.0 there is no distinction anymore between single and multi host collect).
IMPORTANT:
- A STAT collect for a host is independent of any other, so it can be stopped and/or restarted at any time, independent of other collects.
- Your collect options saved into special script files with a name based on the "Collect Base Name". Using customized names you may pre-load a different set of options, according to your needs.
- You may start a collect on-line from your browser, or you can make a start script, to be run by hand, via cron, as a batch job, etc.
Main Steps |
There are 4 main points in starting a STAT Collect:
- choose a host name(s)
- set collect attributes (title, id, etc.)
- choose collected statistics
- start now, or prepare a script for manual/delayed execution
1. Host name(s)
Since version 8.0 you choose your host(s) first. You may setup a list of frequently used host names on the 'Preferences' page. This list as well all other used host names are kept via browser cookies. Before you start any STAT collect, for each given host name, dim_STAT will indicate the status of your host's STAT-service by LED color. I hope it avoids potential misconfiguration issues for both new and experienced users. For now there are 3 LED colors:
- Red: the host is not running STAT-service on the default port, or the host is inaccessible from the network, or the host is down.
- Orange: the host is running STAT-service but an older version.
- Green: ok! STAT-service is running and has all required features.
NOTE: since v.8.0, STAT-service has a new 'stat publish' feature. Using this, the application knows exactly what kind of STATs you can or can't collect from any given host. It protects you from choosing the wrong or unavailable data.
2. Set Collect Attributes
Collect BaseName - all selected options are saved in a special start script. The name of this script is composed of BaseName + some context extentions. When you start a new collect, the next time you may pre-load previously selected options by giving the previous BaseName and clicking on "Preload" (by default the last given BaseName is stored using a cookie).
ID - all data in the database referenced to this ID. The ID is not assigned automatically, to give you a choice to use personalized range numbers (your project id; etc.).
Title - the title description you give for starting the collect.
Time Interval - how frequent (in secs.) you want dim_STAT to collect data (the default is 30 seconds, which is right in most cases)
Client Log File - the name of a file on the "host" that you want to watch. All text lines appended to this file will be automatically be copied into the STAT database and timestamped. While analyzing the collected STATs you can visualize the log messages that correspond with the analyzed interval. This may be very useful to trace auto-starting jobs, night batches, etc. They also give you a simple and fast way to find the correct time position during data analysis, like "show N minutes before/after/around a selected message".
STAT-service Port - the port number on which STAT-service is running. By default the tool will use the port number given during installation and it's a good practice to use the same port on every host.
3. Choose Statistics
Simply select all statistics you want to collect. Help bullets show a full description of each STAT (if you have JavaScript enabled in your web browser). Better be selective, probably you don't need everything.
A good set of STATs to start with:
- VMSTAT
- MPSTAT
- IOSTAT
- netLOAD (avoid using 'netstat')
- ProcLOAD
These STATs will give you a good overview of the resource utilization on your hosts. Once you have analyzed them, you may go more in-depth and fine-tune the selected STATs.NOTE: all "official" Add-Ons are installed by default in each dim_STAT database, BUT! not enabled by default in the STAT-service! On the host side only surely working stats are enabled! Be sure to check /etc/STATsrv/access file on each server before you're starting any collect! :-)
For example: if you're needing to collect "vxstat" data, and you know a VxVM is installed and running on this host - just uncomment the VXSTAT line in your /etc/STATsrv/access file and things will work!...
4. Start Mode
[ Make Start script only ] - don't start the collect, just create a script
[ Start Now! ] - start the new STAT collect right now
[*] Show Debug output - in case you want to see debug messages about the collect startup
Few screenshots... |
Select Host(s) |
You may see here several servers:
- neel, fourrier - Solaris hosts running upgraded STAT-service
- localhost - Linux box, upgraded STAT-service
- sting - Solaris host, old STAT-service
- fudji - Solaris host, powered off
I select neel, fourrier, localhost and sting and click on [Continue] button...
Choose STATs |
The hosts are chosen, let's select the STATs to collect.
Some remarks about these hosts:
- Linux stats are not proposed for 'green' Solaris hosts
- Solaris stats are absent for 'green' Linux hosts
- for any 'green' host, not configured or disabled stats are absent
- the 'orange' host — sting — has all its stats present, but as it was from before v.8.0, it's up to you to remember which commands will run on it or not
Choose STATs, next |
Load collect from output files |
If you cannot collect data directly from your hosts and all you have is a set of statistics output files, then you may still download them via the Web interface as a STAT collect and analyze later. Just fill the required parameters and of you go. However, if your output files are representing a big volume, it may take much more time to load, and your browser may simply timeout and loose the connection. And you'll never see the final result. In such cases, a better solution is to use EasySTAT (simplified) or BatchLOAD (for more experimented users). See the following sections for more details.
Standalone configuration |
Before you think about collecting your stats via some kind of scripts, don't forget about the possibility of a "standalone" dim_STAT. There is absolutely no restriction to:
- install dim_STAT on a host
- start STAT-service on the same host
- collect data from that host into dim-STAT on the same host
- and be aware, on a 4 CPU machine (which is relatively small server) a collect with a 20 second interval (vmstat + mpstat + iostat + psSTAT + netLOAD), will generate only 0.2% CPU usage. (Yes !!)
The CPU usage of dim_STAT for collecting data is very low. However, during data analysis or when doing export/import/etc. actions, CPU utilization is very high.
So, don't forget about this simple solution: install dim_STAT on the same host you want to collect from, collect locally all the data you need, and then backup the whole database, copy it onto another machine and analyze it there. Alternatively, in the case of a benchmark, keep the data on the same server, but take care that you're not doing any analysis at the same time as you're running your testruns.
<< | [up] | >> |